Re: [OMPI devel] Info about ORTE structure

Ralph Castain Thu, 11 Jun 2015 11:09:35 -0400 (EDT)

> On Jun 11, 2015, at 1:13 AM, Federico Reghenzani 
> <[email protected]> wrote:
> 
> Just some other questions on ras functions:
> - we found this old thread 
> <http://www.open-mpi.org/community/lists/users/2007/04/2974.php> in which you 
> said that orte doesn't allow to spawn processes outside ras allocation (e.g. 
> with MPI_Comm_spawn). Is it still  like that? So, allocate() function is 
> called only once time?


Not really - you can use an MPI_Info key to “add-host” or “add-hostfile”

> - where can we found the number of requested processes for the task? We 
> searched in orte_job_t passed to allocate() but we cannot find it. We don't 
> know if take it from argv ("mpirun -np 4 ...") it's a good strategy.

It is in the orte_job_t:

    /* number of procs in this job */
    orte_vpid_t num_procs;


> 
> 
> We really appreciate your help guys :)
> 
> Cheers,
> Federico Reghenzani
> 
> 2015-04-13 21:08 GMT+02:00 Ralph Castain <[email protected] 
> <mailto:[email protected]>>:
> Yes - but the processes must stay in the same location
> 
>> On Apr 13, 2015, at 12:02 PM, Federico Reghenzani 
>> <[email protected] 
>> <mailto:[email protected]>> wrote:
>> 
>> Thank you.
>> 
>> And, to workaround, is it possible to temporary suspend processes on a node 
>> and later resume it (requested by RM)?  I saw in the code that orted can 
>> receive SIGTSTP and SIGCONT to suspend/resume processes.
>> 
>> 
>> Cheers,
>> Federico Reghenzani
>> 
>> 2015-04-10 16:58 GMT+02:00 Ralph Castain <[email protected] 
>> <mailto:[email protected]>>:
>> I’m afraid not. The MPI job would not be very happy to suddenly lose some 
>> nodes during execution, and relocating MPI processes during execution is 
>> something we don’t currently support.
>> 
>> There is work underway to integrate the RM more fully into that procedure so 
>> it could tell the MPI job to checkpoint, wait until that completed, 
>> terminate the job, and then fast-restart it on the new nodes - but that 
>> isn’t here yet.
>> 
>> 
>>> On Apr 10, 2015, at 7:54 AM, Federico Reghenzani 
>>> <[email protected] 
>>> <mailto:[email protected]>> wrote:
>>> 
>>> The RM can ask for deallocation of some nodes?
>>> 
>>> For example, mpirun asks to the RM which resources are available (let 
>>> node1, node2, node3) and spawns orted in the nodes. After some time during 
>>> the elaboration, can the RM ask to deassign node3 or  reassign jobs on 
>>> node3 to node4?
>>> 
>>> Cheers,
>>> Federico Reghenzani
>>> 
>>> 2015-03-26 18:09:22 GMT+06:00 Artem Polyakov <artpol84_at_[hidden]>:  
>>> 
>>> P.S. also check ESS (orte/mca/ess) for environment setup. 
>>> 2015-03-26 18:06 GMT+06:00 Artem Polyakov <artpol84_at_[hidden]>: 
>>> > 
>>> > 2015-03-26 17:58 GMT+06:00 Gianmario Pozzi <pozzigmario_at_[hidden]>: 
>>> > 
>>> >> Hi everyone, 
>>> >> I'm an italian M.Sc. student in Computer Engineering at Politecnico di 
>>> >> Milano. 
>>> >> 
>>> >> My team and I are trying to integrate OpenMPI with a real time resource 
>>> >> manager written by a group of students named BBQ ( 
>>> >> http://bosp.dei.polimi.it/ <http://bosp.dei.polimi.it/> ). We are 
>>> >> encountering some troubles, though. 
>>> >> 
>>> >> Our main issue is to understand how ORTE interacts with the resource 
>>> >> manager, which parts of the code (if any) are executed on the "slave" 
>>> >> nodes 
>>> >> and which ones on the "master". 
>>> >> We spent some time looking at the source code but we still have many 
>>> >> doubts. 
>>> >> 
>>> > 
>>> > Hello, 
>>> > check orte/mca/plm and orte/mca/ras 
>>> > PLM - process lifecycle manager 
>>> > RAS - resource allocation subsystem. 
>>> > 
>>> > In RAS mpirun detects under which RM it works and gets the allocation. 
>>> > in PLM spawn of remote processes is done. 
>>> > mpirun spawns orted daemons on the slave nodes and all the rest is done 
>>> > without RM intervention (IMHO). 
>>> > 
>>> > 
>>> >> 
>>> >> Thank you. 
>>> >> 
>>> >> _______________________________________________ 
>>> >> devel mailing list 
>>> >> devel_at_[hidden] 
>>> >> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel 
>>> >> <http://www.open-mpi.org/mailman/listinfo.cgi/devel> 
>>> >> Link to this post: 
>>> >> http://www.open-mpi.org/community/lists/devel/2015/03/17157.php 
>>> >> <http://www.open-mpi.org/community/lists/devel/2015/03/17157.php> 
>>> >> 
>>> > 
>>> > 
>>> > 
>>> > -- 
>>> > Ð¡ Ð£Ð²Ð°Ð¶ÐµÐ½Ð¸ÐµÐ¼, ÐŸÐ¾Ð»Ñ ÐºÐ¾Ð² Ð Ñ€Ñ‚ÐµÐ¼ Ð®Ñ€ÑŒÐµÐ²Ð¸Ñ‡ 
>>> > Best regards, Artem Y. Polyakov 
>>> > 
>>> 
>>> --
>>> Ð¡ Ð£Ð²Ð°Ð¶ÐµÐ½Ð¸ÐµÐ¼, ÐŸÐ¾Ð»Ñ ÐºÐ¾Ð² Ð Ñ€Ñ‚ÐµÐ¼ Ð®Ñ€ÑŒÐµÐ²Ð¸Ñ‡
>>> Best regards, Artem Y. Polyakov
>>> _______________________________________________
>>> devel mailing list
>>> [email protected] <mailto:[email protected]>
>>> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel 
>>> <http://www.open-mpi.org/mailman/listinfo.cgi/devel>
>>> Link to this post: 
>>> http://www.open-mpi.org/community/lists/devel/2015/04/17210.php 
>>> <http://www.open-mpi.org/community/lists/devel/2015/04/17210.php>
>> 
>> _______________________________________________
>> devel mailing list
>> [email protected] <mailto:[email protected]>
>> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel 
>> <http://www.open-mpi.org/mailman/listinfo.cgi/devel>
>> Link to this post: 
>> http://www.open-mpi.org/community/lists/devel/2015/04/17211.php 
>> <http://www.open-mpi.org/community/lists/devel/2015/04/17211.php>
>> 
>> _______________________________________________
>> devel mailing list
>> [email protected] <mailto:[email protected]>
>> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel 
>> <http://www.open-mpi.org/mailman/listinfo.cgi/devel>
>> Link to this post: 
>> http://www.open-mpi.org/community/lists/devel/2015/04/17215.php 
>> <http://www.open-mpi.org/community/lists/devel/2015/04/17215.php>
> 
> _______________________________________________
> devel mailing list
> [email protected] <mailto:[email protected]>
> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel 
> <http://www.open-mpi.org/mailman/listinfo.cgi/devel>
> Link to this post: 
> http://www.open-mpi.org/community/lists/devel/2015/04/17216.php 
> <http://www.open-mpi.org/community/lists/devel/2015/04/17216.php>
> 
> _______________________________________________
> devel mailing list
> [email protected]
> Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel
> Link to this post: 
> http://www.open-mpi.org/community/lists/devel/2015/06/17491.php

Re: [OMPI devel] Info about ORTE structure

Reply via email to